Assessing Bank Resilience with Regression and Financial Data
Summary
The document presents a student’s proposed comparison of resilience among large Canadian and U.S. banks. The suggested approach regresses quarterly return on assets on national inflation, unemployment, and GDP, then treats weaker relationships with those macroeconomic variables as evidence that a bank is less affected by its environment. The sample spans multiple decades and includes five banks from each country. The student reports similar average fit statistics for the two groups and frequent p-values above a conventional threshold, and asks how to improve the analysis. These results do not establish which country’s banks are more resilient: fit measures describe the chosen model’s explanatory performance, and they do not directly measure resilience. The document supplies a research question and an initial design, but no answers or validated conclusions. It leaves important choices unresolved, including how to define resilience, compare banks and crisis periods, and account for differences in bank characteristics and time-series behavior.
Key ideas
- The proposed study compares large Canadian and U.S. banks using return on assets and national macroeconomic variables.
- The student interprets weaker regression relationships as evidence of greater bank resilience.
- Similar reported fit statistics do not provide clear support for the stated country comparison.
- The document asks for methodological guidance but does not offer a tested alternative or conclusion.
- A defensible comparison depends on how resilience and crisis periods are defined.
Tags
Full text
# How to analyse the resilience of banks during financial crises using linear regression and other statistical methods? # How to analyse the resilience of banks during financial crises using linear regression and other statistical methods? I am a student in finance and have to work on a project for the semester. I have to study the difference of resilience during financial crises between the 5 biggest US banks and the 5 biggest Canadian banks. I know that Canadian banks are praised for their resilience and strong regulations. That is why I chose the following hypothesis that I want to prove: Canadian banks are more resilient than U.S. banks during periods of financial turmoil. My teacher wants me to use statistical methods to support my hypothesis, these methods include linear regression, variances analysis, means analysis, etc. However, my first results are really disappointing and I wanted to have your point of view regarding what I have been doing until now. I decided to start with the linear regression on Excel to see if I get results that could support my hypothesis. I am currently studying 10 banks, 5 Canadian and 5 American. Canadian banks: - TD Bank - Bank of Montreal (BMO) - Royal Bank of Canada (RBC) - The Bank of Nova Scotia - The Canadian Imperial Bank of Commerce (CIBC) American banks: - Bank of America - Wells Fargo - JPMorgan - Citigroup - US Bancorp For my first linear regression, I took one dependent variable (y) and three independent variables (x1), (x2), and (x3). For my dependent variable (y), I chose to use the Return on Asset (ROA) of each banks from Q1 1997 to Q4 2022. Which represent 104 observations which I think is enough for my linear regression. For my independent variables (x1), (x2), and (x3), I used: - (x1): national inflation rates (%) - (x2): national unemployment rates (%) - (x3): national GDP My data sources are Capital IQ for financial data and OECD. My idea behind this regression is that, is there is a lower correlation between the bank's ROA and the three independent variables, it means that the bank is less impacted by the global financial and economical situation of the country and therefore, more resilient. On the other side, if there is a higher correlation, it means that the bank is more easily impacted, and therefore, less resilient. However, I am very disappointed by my results. The average R and R² of Canadian banks are respectively 0.43 and 0.20 and for U.S. banks they are around 0.45 and 0.22. There is a slight difference but nothing crazy. I was expecting a bit more to be honest. And for the P-values, they are often above 0.05 which, if I am not wrong, is not great. (Should be lower than 0.05) Can I have your thoughts on my subject? Do you think I am missing something or just not during it the right way? I would be glad to have your reviews and suggestions. It would greatly help me. Thank you very much!
Shown in full with attribution under the source's licence. Licence: CC BY-SA 4.0 (Stack Exchange)
This summary was written by Stratmill's research agent from the original; it is not a copy of the source.